Social Network Analysis

Social network analysis (SNA) is the study of social structures using graph theory: it represents entities—people, organizations—as nodes, and the relationships between them—friendship, collaboration, communication—as edges. A caveat is in order: this is the structural analysis of networks, not the management of profiles on social platforms. Its aim is to understand how each actor's position in the network shapes the flow of information and influence.

Centrality metrics

The heart of SNA are the measures of centrality, which quantify a node's importance. Degree counts how many connections it has. Betweenness measures how often a node acts as a bridge on the shortest paths between two others—that is, its role as a bottleneck of flow. Closeness reflects how quickly a node can reach the rest. And eigenvector centrality scores a node according to the importance of its neighbors: it counts not just how many connections it has, but their quality. The famous PageRank is a relative of this last one.

Communities, bridges and small-world

On that basis, structural phenomena are studied. Community detection looks for densely connected groups within the network. Homophily describes the tendency to relate to those who resemble us. Weak ties and structural holes—concepts linked to Mark Granovetter and Ronald Burt—explain why bridges between groups bring new information. And the small-world effect, formalized by Duncan Watts and Steven Strogatz in 1998, captures that in many real networks any two nodes are separated by surprisingly short chains.

Where it comes from

Its roots lie in Jacob Moreno's sociometry, which in the 1930s introduced the sociogram—the diagram of dots and lines for relationships between people—and in the graph theory that provides its mathematical machinery. The term “social network” began to be used systematically in 1954, and the field expanded strongly from the 1970s onward.

Its connection to AI

Today SNA intersects with artificial intelligence. Large-scale graph analysis draws on current computing power, and above all graph neural networks have emerged, architectures that learn representations of nodes by encoding the network's structure. They are applied to tasks such as link prediction—anticipating which connections will appear—or node classification, with uses in recommendation, information diffusion and epidemiology.